Do Dollars Drive Decisions

Michael ran the loan department at a regional bank. His team approved thousands of loans annually—personal loans, mortgages, business lines of credit. But recently, complaints surfaced. Lower-income applicants claimed they received worse loan terms than wealthier borrowers with similar credit profiles. Michael needed to know if this was true—and if so, whether it was justified by risk or evidence of bias in his lending process.

Michael hired a data analyst to examine loan approvals across income levels. The analyst studied interest rates, loan amounts, credit grades, debt-to-income ratios, and credit utilization across high-income and low-income applicants. The goal was simple: determine if income influenced loan terms, and if so, whether that influence was based on objective financial risk or unconscious bias.

The data showed a clear difference. High-income applicants received an average interest rate of 10.90%. Low-income applicants got 11.99%—a gap of 1.09 percentage points. This wasn't a rounding error. Over the life of a $50,000 loan, that difference meant thousands of dollars. The question was: did the data justify this gap, or was his bank treating people unfairly?

Michael looked at loan amounts. Here, the picture was different. Income and loan size showed only a weak correlation of 0.37. High-income applicants didn't automatically get bigger loans. This surprised him—it meant income alone wasn't driving loan amounts. Other factors like creditworthiness, debt obligations, and collateral mattered more.

Credit grades told another story. High-income applicants received more A and B grades. Low-income applicants clustered in C, D, and E grades. This pattern raised red flags. Were these grades reflecting true risk, or were loan officers unconsciously favoring wealthier applicants? Michael needed to dig deeper into the underlying risk factors.

The analyst examined debt-to-income ratios and credit utilization—two hard measures of financial risk. High-income borrowers had lower debt-to-income ratios and used less of their available credit. They objectively posed lower financial risk. This data justified the lower interest rates. Wealthier applicants weren't getting preferential treatment—they were genuinely safer bets.

But Michael still worried about the credit grading system. Even though high-income borrowers showed better financials, were the grade differences proportional to the risk differences? Or were loan officers giving A grades to high earners too easily while being overly harsh on lower earners with decent credit? The overrepresentation of high grades among wealthy applicants suggested potential grading bias, even if the interest rate gap was justified.

Michael ran scenarios. If he removed income as a visible data point and had loan officers grade applicants based purely on debt-to-income ratio, credit score, and payment history, would the grades change? He suspected they might. Seeing a six-figure salary created unconscious positive bias. Seeing a modest income created unconscious skepticism—even when the underlying numbers were solid.

Michael restructured his loan approval process. Interest rates would continue to reflect risk—that was fair and financially sound. But he blinded loan officers to applicant income during the initial credit grading phase. Officers saw only credit scores, debt ratios, payment histories, and employment stability—not salary figures. This removed the unconscious bias trigger while preserving risk-based decision making.

He also standardized the grading criteria. Instead of subjective assessments, grades were now algorithmically assigned based on measurable risk factors: debt-to-income ratio thresholds, credit score ranges, delinquency history. If two applicants had the same risk profile, they got the same grade—regardless of income. Human judgment still mattered for edge cases, but the system started objective.

Michael tested the new process for six months. The interest rate gap persisted—10.9% vs 12.0%—because the underlying risk difference was real. But the credit grade distribution shifted. More low-income applicants with strong financials received A and B grades. Fewer high-income applicants with mediocre debt ratios got automatic As. The grades now reflected risk more accurately, not income levels.

Today, Michael's bank lends based on data, not demographics. Income still correlates with loan terms, but only because it correlates with financial stability. The lending process is defensible, transparent, and fair. Complaints dropped. Regulatory reviews found no bias. Michael learned that treating different groups differently isn't discrimination—as long as the difference is justified by objective risk, not subjective assumptions about who deserves better terms.